Prediksi Konsentrasi Emisi NOx pada Pembangkit Listrik Tenaga Uap Menggunakan Model Grey-Box Berbasis Functional Data Analysis

Rohmah, Aulia Afifatur (2026) Prediksi Konsentrasi Emisi NOx pada Pembangkit Listrik Tenaga Uap Menggunakan Model Grey-Box Berbasis Functional Data Analysis. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221055-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (28MB) | Request a copy

Abstract

Pembangkit Listrik Tenaga Uap (PLTU) sebagai salah satu sumber utama energi listrik di Indonesia menghasilkan emisi Nitrogen Oksida (NOₓ) yang berdampak negatif terhadap kesehatan dan lingkungan, sehingga konsentrasinya dibatasi oleh regulasi pemerintah. Oleh karena itu, diperlukan model prediksi konsentrasi emisi NOₓ yang akurat sekaligus memiliki interpretasi fisik yang jelas guna mendukung upaya pengendalian emisi. Penelitian ini mengembangkan model prediksi emisi NOₓ menggunakan pendekatan grey-box berbasis Functional Data Analysis (FDA) yang mengintegrasikan pengetahuan mekanisme pembentukan NOₓ dengan data historis operasi PLTU. Data NOₓ dengan resolusi per menit dari Continuous Emission Monitoring System (CEMS) digunakan sebagai variabel respon, dengan aliran massa batu bara, aliran massa udara primer, dan aliran massa udara sekunder sebagai kovariat yang direpresentasikan melalui elemen fundamental. Pemodelan dilakukan menggunakan regresi fungsional function-on-function berbasis boosting functional regression, dengan dua pendekatan temporal, yaitu concurrent yang merepresentasikan pengaruh fungsi dasar elemen fundamental terhadap emisi NOₓ pada waktu yang sama dan short-term feed-forward yang mempertimbangkan efek historis. Kombinasi elemen fundamental terbaik pada model grey-box concurrent functional regression selanjutnya dimodelkan menggunakan model grey-box short-term feed-forward functional regression dengan mempertimbangkan efek historis 5 menit hingga 24 jam sebelumnya. Hasil evaluasi model menunjukkan bahwa model grey-box short-term feed-forward functional regression dengan efek historis 24 jam memberikan kinerja prediksi terbaik dengan nilai RMSE, MAPE, dan sMAPE terendah dibandingkan model grey-box concurrent functional regression. Sejalan dengan hasil tersebut, rolling-window cross-validation menunjukkan bahwa model short-term feed-forward dengan efek historis 24 jam memiliki struktur model yang stabil dan kemampuan generalisasi yang cukup baik. Selain itu, evaluasi horizon prediksi menunjukkan bahwa model memberikan kinerja prediksi yang cukup baik hingga sekitar 11–13 hari, sehingga relevan untuk mendukung perencanaan operasional dan pengendalian emisi NOₓ di PLTU.
==================================================================================================================================
Coal-Fired Power Plants (PLTU), as one of the main sources of electricity generation in Indonesia, produce Nitrogen Oxides (NOₓ) emissions that have negative impacts on human health and the environment. Therefore, their concentrations are regulated by government regulations. Consequently, accurate NOₓ emission prediction model with physical interpretability is required to support emission control efforts. This study develops a NOₓ emission prediction model using a grey-box approach based on Functional Data Analysis (FDA), which integrates knowledge of NOₓ formation mechanisms with historical operational data from coal-fired power plant. Minute-resolution NOₓ data obtained from the Continuous Emission Monitoring System (CEMS) are used as the response variable, while coal, primary air, and secondary air mass flow rate are used as covariates represented through fundamental elements. The modeling is performed using function-on-function functional regression based on boosting functional regression with two temporal approaches: a concurrent model that represents the influence of fundamental element functions on NOₓ emissions at the same time, and a short-term feed-forward model that accounts for historical effects. The best combination of fundamental elements identified in the grey-box concurrent functional regression model is subsequently modeled using a grey-box short-term feed-forward functional regression framework, considering historical effects ranging from 5 minutes to 24 hours. Model evaluation results indicate that the grey-box short-term feed-forward model with a 24-hour historical effect achieves the best predictive performance, as evidenced by the lowest RMSE, MAPE, and sMAPE values compared to the grey-box concurrent model. Consistent with these results, rolling-window cross-validation demonstrates that the short-term feed-forward model with a 24-hour historical effect exhibits a stable model structure and satisfactory generalization capability. Furthermore, prediction horizon evaluation shows that the model maintains good predictive performance for up to approximately 11–13 days, indicating its relevance for supporting operational planning and NOₓ emission control in coal-fired power plant.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	CEMS, Coal-Fired Power Plant, Grey-Box, FDA, NOx, CEMS, FDA, Grey-Box, NOx, PLTU,
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Aulia Afifatur Rohmah
Date Deposited:	29 Jan 2026 09:42
Last Modified:	29 Jan 2026 09:42
URI:	http://repository.its.ac.id/id/eprint/131145

Actions (login required)

View Item